Decision-Relevance and Irrelevant Questions: Deviation in Strategy-Proof Matching Mechanisms
Draft available upon request
Abstract
Strategy-proofness rules out profitable manipulation, but it does not imply that every deviation from truthful reporting is strategically meaningful. Many matching mechanisms ask agents to submit complete rank-order lists, even though only a subset of the reported comparisons can affect an agent's own assignment. This paper introduces decision-relevance as a way to distinguish report changes that can alter allocations from report changes that are allocatively inconsequential.
Theoretically, I characterize the zero-incentive region of the report space under canonical strategy-proof assignment mechanisms and show how informational redundancy arises in Serial Dictatorship, Random Serial Dictatorship, and Deferred Acceptance. Empirically, I use experimental evidence from rank-order-list reporting to separate allocatively costless deviations from deviations that can carry strategic cost. In the multi-round data, deviations are more frequent on decision-irrelevant comparisons, and a substantial share of misreporting rounds are pure Type I deviations: the submitted list differs from truth but is allocation-equivalent to truth. The results suggest that observed deviations under strategy-proof mechanisms need not reflect misunderstanding alone; they can also reflect the mechanism's own over-elicitation of payoff-irrelevant information.
